Maker Pro
Maker Pro

Grounding: should Vss = chassis?

R

Richard

Jan 1, 1970
0
I'm building a comm device that'll be housed in a metal chassis and
rack-mounted. The question at-hand is whether the PCB ground should be
tied to the chassis or left floating.

Is there any benefit to tying the PCB ground to the chassis ground?
Conversely, are there drawbacks? From my reading, I'm not sure that
it'd reduce EMI emissions, and I'm concerned that it may expose Vss to
external interference. There doesn't seem to be a safety issue, but are
there certification mandates (e.g., UL)?

FYI, the incoming line power will be appropriately grounded to the
chassis for safety. And the metal shells of jacks (RJ-45) would be
grounded to the chassis, not Vss.
 
J

John Larkin

Jan 1, 1970
0
I'm building a comm device that'll be housed in a metal chassis and
rack-mounted. The question at-hand is whether the PCB ground should be
tied to the chassis or left floating.

Is there any benefit to tying the PCB ground to the chassis ground?
Conversely, are there drawbacks? From my reading, I'm not sure that
it'd reduce EMI emissions, and I'm concerned that it may expose Vss to
external interference. There doesn't seem to be a safety issue, but are
there certification mandates (e.g., UL)?

FYI, the incoming line power will be appropriately grounded to the
chassis for safety. And the metal shells of jacks (RJ-45) would be
grounded to the chassis, not Vss.

I'd vote to bolt the PCB to the chassis in as many places as possible,
which usually turns out to be all the mounting holes. The mounting
holes on the PCB should be nailed right into the ground plane. That's
what we do with all our high-speed (and even low-level analog) stuff.

If you do anything else, you are explicitly hoping that the PCB ground
plane will be at different RF potentials from the nearby chassis and
the RJ-45 common mode RF level, and that doesn't make sense.

John
 
G

Genome

Jan 1, 1970
0
Richard said:
I'm building a comm device that'll be housed in a metal chassis and
rack-mounted. The question at-hand is whether the PCB ground should be
tied to the chassis or left floating.

Is there any benefit to tying the PCB ground to the chassis ground?
Conversely, are there drawbacks? From my reading, I'm not sure that
it'd reduce EMI emissions, and I'm concerned that it may expose Vss to
external interference. There doesn't seem to be a safety issue, but are
there certification mandates (e.g., UL)?

FYI, the incoming line power will be appropriately grounded to the
chassis for safety. And the metal shells of jacks (RJ-45) would be
grounded to the chassis, not Vss.

There is no simple answer.

You have to look at your application, buy the relevant standards, read them
and apply according to your own interpretation.

DNA
 
B

Bob

Jan 1, 1970
0
Richard said:
I'm building a comm device that'll be housed in a metal chassis and
rack-mounted. The question at-hand is whether the PCB ground should be
tied to the chassis or left floating.

Don't let it float away 'till something breaks down. tie PCB Vss to chassis
with ~10 Ohms or so.
Is there any benefit to tying the PCB ground to the chassis ground?
Conversely, are there drawbacks?

Try to remove "ground" from your vocabulary. Everything is differential.

Are there any single-ended signals going in or out of the box? They usually
require special attention if pickup or ESD are possible. Figure out where
the "reference node" of each input or output is and treat it as a signal. Of
course, if all your connectors are RJ-45 (transformer coupled), you probably
don't have anything to worry about.
From my reading, I'm not sure that
it'd reduce EMI emissions, and I'm concerned that it may expose Vss to
external interference. There doesn't seem to be a safety issue, but are
there certification mandates (e.g., UL)?

That depends on to whom you are trying to sell it.
FYI, the incoming line power will be appropriately grounded to the
chassis for safety. And the metal shells of jacks (RJ-45) would be
grounded to the chassis, not Vss.

"bonded" to the chassis is a better term.
 
R

Richard

Jan 1, 1970
0
John said:
I'd vote to bolt the PCB to the chassis in as many
places as possible, which usually turns out to be
all the mounting holes. The mounting holes on the
PCB should be nailed right into the ground plane.
That's what we do with all our high-speed (and even
low-level analog) stuff.

Now, I understood that signals shouldn't be tied together at multiple
points (IIRC, because of the differences in potential that can exist
between the points) - is there an exception to this rule for linking
ground planes?

If you do anything else, you are explicitly hoping
that the PCB ground plane will be at different RF
potentials from the nearby chassis and the RJ-45
common mode RF level, and that doesn't make sense.

Perhaps the Dumb-Question-of-the-Day, but why is that a Bad Thing? Does
the difference in RF potential have a negative effect on the PCB signals
that wouldn't exist if they were coupled? I mean, there will always be
a differential between the PCB and some part of its surroundings - what
dictates where the coupling effort stops?
 
J

John Larkin

Jan 1, 1970
0
Now, I understood that signals shouldn't be tied together at multiple
points (IIRC, because of the differences in potential that can exist
between the points) - is there an exception to this rule for linking
ground planes?

It's not a rule, it's a superstition.
Perhaps the Dumb-Question-of-the-Day, but why is that a Bad Thing? Does
the difference in RF potential have a negative effect on the PCB signals
that wouldn't exist if they were coupled? I mean, there will always be
a differential between the PCB and some part of its surroundings - what
dictates where the coupling effort stops?

Well, if the PCB is flapping around at umpteen MHz relative to a metal
enclosure that may be a fraction of an inch away, it will have signals
induced into its traces, which may or might not matter for a
particular situation. But there's certainly no advantage. The box is a
nice farady cage protecting the PCB from external nasties (and from
radiating as well) but that's lost if the PCB has a high impedance
relative to the box it's in, and the PCB has wire connections to the
outside world where RF can enter and leave, and nice ESD zaps can pump
the PCB ground plane. The box stops being a friend and become the
enemy.

Single-point grounding is sort of an audio myth; at high frequencies,
everything - including those nice star runs to the single-point thing
- is an inductor.


John
 
K

Kevin Aylward

Jan 1, 1970
0
John said:
I'd vote to bolt the PCB to the chassis in as many places as possible,
which usually turns out to be all the mounting holes.

Note that if this was an audio mixing desk, this would be suicide. In
these cases, decent pro desks usually have *everything* isolated from
chassis, with only one connectable lead to connect the electronics to
the chassis.

Kevin Aylward
[email protected]
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.
 
K

Kevin Aylward

Jan 1, 1970
0
John said:
It's not a rule, it's a superstition.

Not for audio.
Well, if the PCB is flapping around at umpteen MHz relative to a metal
enclosure that may be a fraction of an inch away, it will have signals
induced into its traces, which may or might not matter for a
particular situation. But there's certainly no advantage. The box is a
nice farady cage protecting the PCB from external nasties (and from
radiating as well) but that's lost if the PCB has a high impedance
relative to the box it's in, and the PCB has wire connections to the
outside world where RF can enter and leave, and nice ESD zaps can pump
the PCB ground plane. The box stops being a friend and become the
enemy.

Single-point grounding is sort of an audio myth;

No way. With all due respect here, your mistaken.
at high frequencies,
everything - including those nice star runs to the single-point thing
- is an inductor.

High frequencies are not the problem with audio. Its usually ground hum
and dc offset switching clicks. With a sloppy ground connected
everywhere, you are almost certainly *guaranteed* to have hum problems
and offset clicks. My own power amp needs the ground lead from its
capacitor centre point removed from the chassis to eliminate hum. Its a
fact, I can generate the effect at will, so there is nothing
superstitious about it at all. If the back plate cannons were grounded
separately, along with the amp and meters etc, etc, there would be no
practical way to lift the ground. Ground loops are a never ending
nightmare in audio.

Kevin Aylward
[email protected]
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.
 
A

Active8

Jan 1, 1970
0
It's not a rule, it's a superstition.

i was (superstition) with a switcher i had with the output filter ground
tied to the regulator chip's MOSFET ground with a trace. damn noise had
the MOSFET stuck full on, i guess. running two wires for each ground to
the bench supply's ground fixed it. weird, 20kHz switching. too low for
ground bounce. must have been the circulating current. but such a short
trace! i never did analyze that cause/effect.

regards,
mike
 
R

R.Legg

Jan 1, 1970
0
Perhaps the Dumb-Question-of-the-Day, but why is that a Bad Thing? Does
the difference in RF potential have a negative effect on the PCB signals
that wouldn't exist if they were coupled? I mean, there will always be
a differential between the PCB and some part of its surroundings - what
dictates where the coupling effort stops?

Wi'leenks? RF don' need no steenkeen wi'leenks!

http://www.analog.com/ see AN214,AN345, AN346, AN347 and AN358

http://www.lundahl.se/pdfs/papers/groundin.pdf

Perhaps your local technical college or university library still has
something by Ralph Morrison.

RL
 
J

John Larkin

Jan 1, 1970
0
Not for audio.


No way. With all due respect here, your mistaken.


High frequencies are not the problem with audio.

Sometimes they are, as when RF gets into low-level stuff, gets
rectified by junctions, and becomes noise or DC offset. Lots of cheap
telephones, for example, don't work right near radio/TV stations.
Its usually ground hum
and dc offset switching clicks. With a sloppy ground connected
everywhere, you are almost certainly *guaranteed* to have hum problems
and offset clicks. My own power amp needs the ground lead from its
capacitor centre point removed from the chassis to eliminate hum. Its a
fact, I can generate the effect at will, so there is nothing
superstitious about it at all. If the back plate cannons were grounded
separately, along with the amp and meters etc, etc, there would be no
practical way to lift the ground. Ground loops are a never ending
nightmare in audio.

Hum (and microvolt DC offsets in things like thermocouple amps) can be
handled locally without enforcing a star grounding system. I make NMR
gradient amps, 10 to 100-amp wideband current sources; they must have
sub-PPM hum levels or I catch hell. Audio is gross compared to this
stuff. I ground everything to the chassis to minimize RF and
digital-analog crosscoupling, and handle all low-level stuff against
the local ground. As long as you accept that the "ground" plane has AC
and DC gradients, it's not hard to work around. These things are never
"nightmares"; we design them and they work.

John
 
K

Kevin Aylward

Jan 1, 1970
0
John said:
Sometimes they are, as when RF gets into low-level stuff, gets

Ho humm...not in the context that I am obviously discussing.

Hum (and microvolt DC offsets in things like thermocouple amps) can be
handled locally without enforcing a star grounding system.

I was not suggesting that you *everything* has to be connected as a star
ground. What I am saying is that connecting all bits and bobs to the
chassis whenever you feel like it, simple don't work in many cases. If
it does work, you were lucky.

I have just given you a *real* example where if all connections had been
connected to the chassis, I would be sitting here with bloody humming
amp.
I make NMR
gradient amps, 10 to 100-amp wideband current sources; they must have
sub-PPM hum levels or I catch hell. Audio is gross compared to this
stuff.

So what. Pro audio usually has many pieces of equipment all connected
together, all running from different power points. People have spent
days trying to fix grounding problems in studios.

I ground everything to the chassis to minimize RF and
digital-analog crosscoupling, and handle all low-level stuff against
the local ground. As long as you accept that the "ground" plane has AC
and DC gradients, it's not hard to work around. These things are never
"nightmares"; we design them and they work.

I have to disagree. I am very, very familiar with the practical design
of pro audio gear and what actually works in practise. Its not
debatable. Without being able to ground lift, you will have problems,
unless your name is Jesus.


Kevin Aylward
[email protected]
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.
 
K

Kevin Aylward

Jan 1, 1970
0
John said:
Funny, I did two mixer designs this way, and they worked. But the OP
was asking about a digital gadget, which wouldn't be bothered by
millivolt hum loops.

But you were the one to suggest that for audio it was superstition, I
was disagreeing with this view on that issue. My real life experience is
quite the opposite. I bet your most pro audio companies will tell you
right of the bat of the mess-up that had when they built their first
products by earthing everything to chassis.

Kevin Aylward
[email protected]
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.
 
B

Bob

Jan 1, 1970
0
Richard said:
First, thanks for the responses so far. Interesting feedback.

Collectively, comments so far suggest tying Vss to the chassis. I like
the suggestion of using a small resistor, but my reading suggests one
point of connection is better than several. (Comments?)



Hmmm. Till something breaks down? How so? Within a device, shouldn't
it only matter that Vss is a consistently different potential from Vcc -
how does the chassis' potential become significant if it's isolated?

Consistently different potential is the right idea. If you don't make *any*
connection between chasis and PCB Vss (or Vcc or whatever) then you have
built a capacitor. The sum of the leakage currents and the (uncontrolled and
maybe nonlinear) shunt resistance will determine the actaul potential
difference. The better your isolation, the weirder things get.
If the chassis is tied to a PCB signal (be it Vss, Vcc, etc.), won't it
introduce instability by varying the potential on that signal? After
all, isn't this tying a cleaned, regulated power rail directly to a
noise source (since the chassis is tied to the mains ground)?



No, just Ethernet signals, which are differential. There isn't a common
signal ground at play in the design (no RS-232 console port).

I don't think you really have anything to worry about then. Build the thing,
test the thing, and decide for yourself how you want to connect the chassis.
Just remeber to test it thoroughly. Hit it with everything that could ever
happen in the customer's site (and then some). One of my favorite devices is
our "cattle prod". It's a HV PS hooked up to a PVC pipe handle with a wire
wisker sticking out the end. Inside the PVC is an RC (don't remember the
details - we got it from a testing standards doc years ago and now it's
potted inside the PVC ;-) circuit that simulates a human finger. Run the HV
PS up to a few kV and poke the wisker at everything you can (connector pins
and shells, front panel controls and indicators, chassis, EVERYTHING). It
makes a satisfying poping sound and a nice little spark. Lot's of "well
thought out" grounding schemes need re-work after the cattle prod is
used.Signal integrity testing is a whole 'nother thing, but I'll stop now.

Bob
 
J

John Larkin

Jan 1, 1970
0
I ground everything to the chassis to minimize RF and

I have to disagree. I am very, very familiar with the practical design
of pro audio gear and what actually works in practise. Its not
debatable. Without being able to ground lift, you will have problems,
unless your name is Jesus.

OK, if you're interconnecting a bunch of store-bought gear that was
designed by different people, you can expect ground loop problems, and
you'll have to futz around until they go away. But if you're
*designing* a precision instrument and control everything, hard
chassis grounding can be managed nicely, and has a number of
mechanical and ESD/RFI advantages. For an all-digital design,
single-point grounding is not just silly, it's almost meaningless.

The real point is to design it a way that will work, not to follow
"rules" that propagate by hearsay.

John
 
B

Ben Bradley

Jan 1, 1970
0
Even at 50 Hz down to DC, everything, including the chassis, is a
resistor.
Ho humm...not in the context that I am obviously discussing.

Inadvertent RF pickup and rectification are easily controlled in
audio with good RF filtering and shielding. Cheap products (the
mention of which got snipped) generally don't have any
filtering/shielding not required to make the product pass RF emissions
tests.
I was not suggesting that you *everything* has to be connected as a star
ground. What I am saying is that connecting all bits and bobs to the
chassis whenever you feel like it, simple don't work in many cases. If
it does work, you were lucky.

I have just given you a *real* example where if all connections had been
connected to the chassis, I would be sitting here with bloody humming
amp.

I've found AN003 and AN004 from this page to be very well written
articles on where hum in audio systems comes from and how to fix it:

http://www.jensen-transformers.com/apps_wp.html

I see two basic problems: leakage current in power transformers,
and sensitivity of input (and output!) connections to currents into
the ground connection. Of course this applies within a case as well as
when connecting different equipment. Even small currents through a
chassis can cause a voltage drop that gets amplified by a low-level
circuit. The solution is not neccesarily just blind "star grounding"
but seeing where currents go, what is sensitive, and routing grounds
where they (and the currents they carry) won't cause a problem.
 
A

Active8

Jan 1, 1970
0
Even at 50 Hz down to DC, everything, including the chassis, is a
resistor.


Inadvertent RF pickup and rectification are easily controlled in
audio with good RF filtering and shielding. Cheap products (the
mention of which got snipped) generally don't have any
filtering/shielding not required to make the product pass RF emissions
tests.


I've found AN003 and AN004 from this page to be very well written
articles on where hum in audio systems comes from and how to fix it:

http://www.jensen-transformers.com/apps_wp.html

I see two basic problems: leakage current in power transformers,
and sensitivity of input (and output!) connections to currents into
the ground connection. Of course this applies within a case as well as
when connecting different equipment. Even small currents through a
chassis can cause a voltage drop that gets amplified by a low-level
circuit. The solution is not neccesarily just blind "star grounding"
but seeing where currents go, what is sensitive, and routing grounds
where they (and the currents they carry) won't cause a problem.

ok. add small currents/emi near high impedance inputs.

mike
 
A

Active8

Jan 1, 1970
0
Interesting. There's clearly much to learn in this space; for now, the
consensus seems pretty clear that chassis bonding is a good thing, for
reasons that I need to study in my free time. :)

This book by Morrison seems to come up regularly on the topic - do you
have any comments on its usefulness for this scenario, or other
worthwhile titles?
http://www.amazon.com/exec/obidos/tg/detail/-/0471245186/


One nagging question, though... doesn't tying the chassis to an internal
signal add noise on the signal? Or is it a matter of the lesser of two
evils? Are there ways to reduce the impact (perhaps the inline resistor
is the trick)?

The materials I've read are pretty explicit about watching potential on
Vss in particular - such as not sharing a return path for different
supply lines. Tying the chassis to Vss would seem counter to those
warnings, which has me a bit confused.

Vss... was it about ground bounce, where a fast current transition in
the high L bonding wire to a MOSFET source causes a voltage drop across
it, thereby lowering the Vgs to the point where the thing cuts off?

mike
 
K

Kevin Aylward

Jan 1, 1970
0
John said:
OK, if you're interconnecting a bunch of store-bought gear that was
designed by different people, you can expect ground loop problems, and
you'll have to futz around until they go away.

Which is usually the case in pro audio, or any audio for that matter.
But if you're
*designing* a precision instrument and control everything, hard
chassis grounding can be managed nicely, and has a number of
mechanical and ESD/RFI advantages.

But you still have to be very carefull. Potentials on ground plans can
still be such that noise gets in relativly huge amounts. Suppose one is
designing a medical ultrasound imaging system:), and the effective
input noise is << 1nV/rthz at Mhz BW. A Logic switch of 10ma in 10ns in
5nh causing a 5mv spike would be completly hopeless. They is no real way
you can elimitate this sort of noise without making sure the analogue is
single pointed. Even 0.1nH of common impedance will cause severe bother.
For an all-digital design,
single-point grounding is not just silly, it's almost meaningless.

Obviously. 1 Volt noise immunity verses 1uV means that completely
different approaches can be used.


Kevin Aylward
[email protected]
http://www.anasoft.co.uk
SuperSpice, a very affordable Mixed-Mode
Windows Simulator with Schematic Capture,
Waveform Display, FFT's and Filter Design.
 
Top